Intel Burn Test vs. Prime95

I like to use LinX, its based on the same library as Intel Burn Test, which is Linpack, a pretty efficient benchmark used to see how fast supercomputers are in FLOPS, I usually get my chip anywhere from 5 - 8 degrees Celsius hotter then Prime95

If I have to chose I would pick IBT, but I much prefer LinX

LinX

IBT is way faster to test out and better because someones passes prime but fails on IBT. I also heard that Linx (that guy talks about it above) is little better than IBT, I dont have personal experience about that.
Edited by MooMoo - 4/14/10 at 3:47pm

OCCT also runs Linpack, and leaves you with nice graphs.

In Intel Burn Test, you have to change the "Threads" parameter from "Auto" to "8".

I use LinX as a quick method to see if something works or not (20 maxed runs), flags errors way sooner than prime. But I do my final stability checking with 12 hours prime blends.

Sometimes I get errors in prime after 4 to 6 hours that don't flag in LinX...

So it's not one or the other in my case, it's both.

Some people also prefer to use Prime over LinX/IBT because the loadtemps are a bit lower, not everyones OC/cooling can handle that linpack load.

I use the Intel burn test because it is so quick however some people claim it destroys your CPU?

It couldn't possibly destroy the CPU, unless you use stock cooling with their single balance point "contact area"....that they should be sued for. I break the OC rules though and leave thermal management turned on, just in case.

IBT really works it hard though but for me it hasn't been consistent. While it will always work it harder than common alternatives seem to, even with the same settings/same RAM usage, no consecutive runs works it the same. No matter how hard it ends up working it, or not, that it's like that makes comparing runs impossible as a benchark measure.

Does it then retain merit for a simple pass/fail test? I've come up with the impression that it's only those where it works it hardest that can be conclusive, for that could push it to a fail while if it just happens to work it 25% or even just 5% less as per the thermal monitor graphs, then it will be far more likely to pass the tests.

In my case the difference between pushing stop/test/stop/test can be a good 50% thermally and even close to 50% in GFLOPS.

It'd be nice if it were more consistent, but none the less it seems at least OK to kick the legs out of very poor settings early in the game.

What are the thoughts on running other applications while doing a burn test? For example I passed 23 out of 25 runs the other night on max with ITB, and I was doing whatever on it the whole time, browsing etc, but it failed the second I opened uttorrent.

More stable settings/higher voltages don't seem to be so susceptible to IBT GFLOPS wavering while doing other work in it, but lowever voltages that still seem stable can have them vary quite a bit.
Edited by rollemup - 4/15/10 at 10:55am

OCCT failed the very second I opened up CPUZ. Not sure if it was a coincidence.

CPU-Z will tell you the two don't place nice together. I've had similar experiences myself.